GNU bug report logs - #13802
stack overflow in mm-add-meta-html-tag

Previous Next

Full log

Message #11 received at 13802 <at> debbugs.gnu.org (full text, mbox):

From: Stefan Monnier <monnier <at> iro.umontreal.ca>
To: Thien-Thi Nguyen <ttn <at> gnuvola.org>
Cc: 13802 <at> debbugs.gnu.org
Subject: Re: bug#13802: stack overflow in mm-add-meta-html-tag
Date: Sun, 24 Feb 2013 21:04:21 -0500

> I see a "Stack overflow in regexp matcher" error traceable back to
> lisp/gnus/mm-decode.el func ‘mm-add-meta-html-tag’ fragment:

>   (re-search-forward "\
>   <meta\\s-+http-equiv=[\"']?content-type[\"']?\\s-+content=[\"']\
>   text/\\(\\sw+\\)\\(?:\;\\s-*charset=\\(.+\\)\\)?[\"'][^>]*>" nil t)

Hmm... I don't see any obvious reason for a stack overflow unless the
text has some very long lines or a lot of space between elements.

> One idea (untested) is to replace the ".+" (used to match the charset)
> with a more specific pattern.  Perhaps "[^<>]+" or "\\sw+"?

I don't think that would help.  To avoid such overflow, you need to
reduce the backtracking, i.e. reduce the number of cases where two
options are possible according to the simplistic regexp-optimizer.
\s<CHAR> pattern is actually very poor in this respect, because the
optimizer can't know anything about the chars that this matches (since
it depends on text-properties).
The flip side is that replacing \\s- with [ \t\n] might help (this way,
the optimizer will see that the + repetition does not need backtracking
since a char cannot both match a loop iteration and the "after the
loop" content).
Similarly using [^;'\"]+ instead of \\sw+ would help, and maybe replacing
.+ with [^'\"\n]+ would help as well.

> Thinking more systematically, maybe Emacs should add a condition
> ‘stack-overflow/regexp’ (or something like that) such that code can
> ‘condition-case’ for it and try a fallback path.

In reality, such overflow should only ever happen if you have backrefs
in your regexp.


        Stefan

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.